MiniMax M2.5 Python Code Distillation with Verified Solutions

MiniMax M2.5 Code Distilled 14K is a synthetic dataset for code generation created by Madras1. It contains Python coding problems, chain-of-thought reasoning, and execution-verified solutions that passed automated tests.

Use Cases

• Train a code generation model on Python coding problems and their verified correct solutions.
• Analyze the structure of chain-of-thought reasoning provided for each coding problem.
• Fine-tune a model to generate execution-verified solutions from problem descriptions and reasoning steps.

Strengths

• Every solution is execution-verified against test cases in a sandboxed subprocess.
• Includes chain-of-thought reasoning for each coding problem.
• Dataset is specifically distilled from the MiniMax-M2.5 model.

Limitations

• Dataset size and row count are unknown, limiting assessment of scale.
• Scope is limited to Python code generation tasks.
• Synthetic nature may not fully represent real-world coding distributions.

Provenance

Source

Madras1 on Hugging Face.

Collection Method

Created by distilling the MiniMax-M2.5 model; includes synthetic problems, reasoning, and verified solutions.

Freshness

Last updated on March 24, 2026.